There are many fallacies about 64-bit computing. The most common one is that moving to a 64-bit will give you double the performance because it can process twice as much data in one go. This is not true.
A 64-bit architecture actually allows you to directly address a lot more memory (18 exabytes vs 4 gigabytes - an exabyte being a gigabyte with another nine zeroes).
Also, each memory location you address is twice as big, so you can store a very, very large number in it without having to resort to software tricks.
However the actual speed at which the data is processed, does not significantly change - this is still very much related to system architecture and clock speed. From a business or financial systems point of view, the size of the largest number you can store in memory is not terribly important - few companies measure sales or revenues in the quintillions of rands. Being able to manipulate such large numbers is only really important in a few specialised areas such as the scientific, industrial and cryptographic worlds.
So if it's not really faster, and most businesses don't need the very large data registers, what is all the excitement about in 64-bit computing?
It all comes back to the amount of memory you can directly address without having to resort to performance-sapping techniques using virtual memory.
You can now take a large database and analyse it in the same way you do with spreadsheets on your desktop, directly in memory (RAM).
In the 32-bit world, if you had a lot of data to analyse, you would have to manipulate it first and pre-calculate totals for pre-defined hierarchies to give you acceptable
|
response times during the analysis process. Even then, performance remained a big problem.
This was the world of OLAP - online analytical processing which was actually more "off-line". This was the world of data cubes, where you'd decide what questions you'd need to ask, pull the information you'd thought you'd need from the corporate information systems and then build cubes of information that could be analysed to give you answers. And if the cubes did not cover all your requirements, you would have to go back to the IT department and have them rebuild them.
Why all the complexity? Because old-technology business intelligence (BI) tools could not look at the vast amounts of data required to do analysis in one snapshot - they simply couldn't address that much memory directly.
In the 64-bit world, the restrictions of OLAP can be completely eliminated. It is now possible to develop new technologies that are not hindered by hardware constraints, as technologies can take full advantage of the vast amount of memory that is now available.
One of these technologies is called AQL (Associative Query Logic). It was developed in Sweden and is used in the QlikView data analytics tool.
The ease of implementation and use of this technology converts into significant savings from a maintenance, support and training point of view.
It is now possible to deploy a full BI solution in days - not months or years - and reap the benefits immediately. Under these conditions, there is no doubt that the days of traditional OLAP are numbered.
|
